Artificial Intelligence Nanodegree¶
Convolutional Neural Networks¶
Project: Write an Algorithm for a Dog Identification App¶
In this notebook, some template code has already been provided for you, and you will need to implement additional functionality to successfully complete this project. You will not need to modify the included code beyond what is requested. Sections that begin with '(IMPLEMENTATION)' in the header indicate that the following block of code will require additional functionality which you must provide. Instructions will be provided for each section, and the specifics of the implementation are marked in the code block with a 'TODO' statement. Please be sure to read the instructions carefully!
Note: Once you have completed all of the code implementations, you need to finalize your work by exporting the iPython Notebook as an HTML document. Before exporting the notebook to html, all of the code cells need to have been run so that reviewers can see the final implementation and output. You can then export the notebook by using the menu above and navigating to File -> Download as -> HTML (.html). Include the finished document along with this notebook as your submission.
In addition to implementing code, there will be questions that you must answer which relate to the project and your implementation. Each section where you will answer a question is preceded by a 'Question X' header. Carefully read each question and provide thorough answers in the following text boxes that begin with 'Answer:'. Your project submission will be evaluated based on your answers to each of the questions and the implementation you provide.
Note: Code and Markdown cells can be executed using the Shift + Enter keyboard shortcut. Markdown cells can be edited by double-clicking the cell to enter edit mode.
The rubric contains optional "Stand Out Suggestions" for enhancing the project beyond the minimum requirements. If you decide to pursue the "Stand Out Suggestions", you should include the code in this IPython notebook.
Why We're Here¶
In this notebook, you will make the first steps towards developing an algorithm that could be used as part of a mobile or web app. At the end of this project, your code will accept any user-supplied image as input. If a dog is detected in the image, it will provide an estimate of the dog's breed. If a human is detected, it will provide an estimate of the dog breed that is most resembling. The image below displays potential sample output of your finished project (... but we expect that each student's algorithm will behave differently!).
In this real-world setting, you will need to piece together a series of models to perform different tasks; for instance, the algorithm that detects humans in an image will be different from the CNN that infers dog breed. There are many points of possible failure, and no perfect algorithm exists. Your imperfect solution will nonetheless create a fun user experience!
The Road Ahead¶
We break the notebook into separate steps. Feel free to use the links below to navigate the notebook.
- Step 0: Import Datasets
- Step 1: Detect Humans
- Step 2: Detect Dogs
- Step 3: Create a CNN to Classify Dog Breeds (from Scratch)
- Step 4: Use a CNN to Classify Dog Breeds (using Transfer Learning)
- Step 5: Create a CNN to Classify Dog Breeds (using Transfer Learning)
- Step 6: Write your Algorithm
- Step 7: Test Your Algorithm
Step 0: Import Datasets¶
Import Dog Dataset¶
In the code cell below, we import a dataset of dog images. We populate a few variables through the use of the load_files function from the scikit-learn library:
train_files,valid_files,test_files- numpy arrays containing file paths to imagestrain_targets,valid_targets,test_targets- numpy arrays containing onehot-encoded classification labelsdog_names- list of string-valued dog breed names for translating labels
from sklearn.datasets import load_files
from keras.utils import np_utils
import numpy as np
from glob import glob
# define function to load train, test, and validation datasets
def load_dataset(path):
data = load_files(path)
dog_files = np.array(data['filenames'])
dog_targets = np_utils.to_categorical(np.array(data['target']), 133)
return dog_files, dog_targets
# load train, test, and validation datasets
train_files, train_targets = load_dataset('dogImages/train')
valid_files, valid_targets = load_dataset('dogImages/valid')
test_files, test_targets = load_dataset('dogImages/test')
# load list of dog names
dog_names = [item[20:-1] for item in sorted(glob("dogImages/train/*/"))]
# print statistics about the dataset
print('There are %d total dog categories.' % len(dog_names))
print('There are %s total dog images.\n' % len(np.hstack([train_files, valid_files, test_files])))
print('There are %d training dog images.' % len(train_files))
print('There are %d validation dog images.' % len(valid_files))
print('There are %d test dog images.'% len(test_files))
Import Human Dataset¶
In the code cell below, we import a dataset of human images, where the file paths are stored in the numpy array human_files.
import random
random.seed(8675309)
# load filenames in shuffled human dataset
human_files = np.array(glob("lfw/*/*"))
random.shuffle(human_files)
# print statistics about the dataset
print('There are %d total human images.' % len(human_files))
Step 1: Detect Humans¶
We use OpenCV's implementation of Haar feature-based cascade classifiers to detect human faces in images. OpenCV provides many pre-trained face detectors, stored as XML files on github. We have downloaded one of these detectors and stored it in the haarcascades directory.
In the next code cell, we demonstrate how to use this detector to find human faces in a sample image.
import cv2
import matplotlib.pyplot as plt
%matplotlib inline
# extract pre-trained face detector
face_cascade = cv2.CascadeClassifier('haarcascades/haarcascade_frontalface_alt.xml')
# load color (BGR) image
img = cv2.imread(human_files[111])
# convert BGR image to grayscale
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
# find faces in image
faces = face_cascade.detectMultiScale(gray)
# print number of faces detected in the image
print('Number of faces detected:', len(faces))
# get bounding box for each detected face
for (x,y,w,h) in faces:
# add bounding box to color image
cv2.rectangle(img,(x,y),(x+w,y+h),(255,0,0),2)
# convert BGR image to RGB for plotting
cv_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
# display the image, along with bounding box
plt.imshow(cv_rgb)
plt.show()
Before using any of the face detectors, it is standard procedure to convert the images to grayscale. The detectMultiScale function executes the classifier stored in face_cascade and takes the grayscale image as a parameter.
In the above code, faces is a numpy array of detected faces, where each row corresponds to a detected face. Each detected face is a 1D array with four entries that specifies the bounding box of the detected face. The first two entries in the array (extracted in the above code as x and y) specify the horizontal and vertical positions of the top left corner of the bounding box. The last two entries in the array (extracted here as w and h) specify the width and height of the box.
Write a Human Face Detector¶
We can use this procedure to write a function that returns True if a human face is detected in an image and False otherwise. This function, aptly named face_detector, takes a string-valued file path to an image as input and appears in the code block below.
# returns "True" if face is detected in image stored at img_path
def face_detector(img_path):
img = cv2.imread(img_path)
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
faces = face_cascade.detectMultiScale(gray)
return len(faces) > 0
(IMPLEMENTATION) Assess the Human Face Detector¶
Question 1: Use the code cell below to test the performance of the face_detector function.
- What percentage of the first 100 images in
human_fileshave a detected human face? - What percentage of the first 100 images in
dog_fileshave a detected human face?
Ideally, we would like 100% of human images with a detected face and 0% of dog images with a detected face. You will see that our algorithm falls short of this goal, but still gives acceptable performance. We extract the file paths for the first 100 images from each of the datasets and store them in the numpy arrays human_files_short and dog_files_short.
Answer: 98% of the first 100 images in human_files have a detected human face, and 11% of the first 100 images in dog_files have a detected human face.
human_files_short = human_files[:100]
dog_files_short = train_files[:100]
# Do NOT modify the code above this line.
## Test the performance of the face_detector algorithm
## on the images in human_files_short and dog_files_short.
n_human_faces = sum([face_detector(file) for file in human_files_short])
n_dog_faces = sum([face_detector(file) for file in dog_files_short])
print("Percentage of first 100 human images that have a detected human face: {}%".format(n_human_faces))
print("Percentage of first 100 dog images that have a detected human face: {}%".format(n_dog_faces))
Question 2: This algorithmic choice necessitates that we communicate to the user that we accept human images only when they provide a clear view of a face (otherwise, we risk having unneccessarily frustrated users!). In your opinion, is this a reasonable expectation to pose on the user? If not, can you think of a way to detect humans in images that does not necessitate an image with a clearly presented face?
Answer: I think the answer to this question depends upon the user's motivation for using the face detector. In many situations, requiring a clear view of a face is a reasonable expectation to impose upon the user. However, I can also think of many situations where the user would appreciate or even expect a higher level of robustness. If this is the case, then we could probably do much better if we use a face detection model that has been extensively trained on an augmented set of human face images. An augmented set would include images that have been modified in one or more ways, for example:
- image rotations,
- image translations (up, down, left, or right),
- image scaling (stretched or compressed),
- image color balance modifications, and others.
This augmented training process would do a much better job of predicting the presence of a human face in images that do not have a clearly presented face.
We suggest the face detector from OpenCV as a potential way to detect human images in your algorithm, but you are free to explore other approaches, especially approaches that make use of deep learning :). Please use the code cell below to design and test your own face detection algorithm. If you decide to pursue this optional task, report performance on each of the datasets.
## (Optional) TODO: Report the performance of another
## face detection algorithm on the LFW dataset
### Feel free to use as many code cells as needed.
Step 2: Detect Dogs¶
In this section, we use a pre-trained ResNet-50 model to detect dogs in images. Our first line of code downloads the ResNet-50 model, along with weights that have been trained on ImageNet, a very large, very popular dataset used for image classification and other vision tasks. ImageNet contains over 10 million URLs, each linking to an image containing an object from one of 1000 categories. Given an image, this pre-trained ResNet-50 model returns a prediction (derived from the available categories in ImageNet) for the object that is contained in the image.
from keras.applications.resnet50 import ResNet50
# define ResNet50 model
ResNet50_model = ResNet50(weights='imagenet')
Pre-process the Data¶
When using TensorFlow as backend, Keras CNNs require a 4D array (which we'll also refer to as a 4D tensor) as input, with shape
$$ (\text{nb_samples}, \text{rows}, \text{columns}, \text{channels}), $$where nb_samples corresponds to the total number of images (or samples), and rows, columns, and channels correspond to the number of rows, columns, and channels for each image, respectively.
The path_to_tensor function below takes a string-valued file path to a color image as input and returns a 4D tensor suitable for supplying to a Keras CNN. The function first loads the image and resizes it to a square image that is $224 \times 224$ pixels. Next, the image is converted to an array, which is then resized to a 4D tensor. In this case, since we are working with color images, each image has three channels. Likewise, since we are processing a single image (or sample), the returned tensor will always have shape
The paths_to_tensor function takes a numpy array of string-valued image paths as input and returns a 4D tensor with shape
Here, nb_samples is the number of samples, or number of images, in the supplied array of image paths. It is best to think of nb_samples as the number of 3D tensors (where each 3D tensor corresponds to a different image) in your dataset!
from keras.preprocessing import image
from tqdm import tqdm
def path_to_tensor(img_path):
# loads RGB image as PIL.Image.Image type
img = image.load_img(img_path, target_size=(224, 224))
# convert PIL.Image.Image type to 3D tensor with shape (224, 224, 3)
x = image.img_to_array(img)
# convert 3D tensor to 4D tensor with shape (1, 224, 224, 3) and return 4D tensor
return np.expand_dims(x, axis=0)
def paths_to_tensor(img_paths):
list_of_tensors = [path_to_tensor(img_path) for img_path in tqdm(img_paths)]
return np.vstack(list_of_tensors)
Making Predictions with ResNet-50¶
Getting the 4D tensor ready for ResNet-50, and for any other pre-trained model in Keras, requires some additional processing. First, the RGB image is converted to BGR by reordering the channels. All pre-trained models have the additional normalization step that the mean pixel (expressed in RGB as $[103.939, 116.779, 123.68]$ and calculated from all pixels in all images in ImageNet) must be subtracted from every pixel in each image. This is implemented in the imported function preprocess_input. If you're curious, you can check the code for preprocess_input here.
Now that we have a way to format our image for supplying to ResNet-50, we are now ready to use the model to extract the predictions. This is accomplished with the predict method, which returns an array whose $i$-th entry is the model's predicted probability that the image belongs to the $i$-th ImageNet category. This is implemented in the ResNet50_predict_labels function below.
By taking the argmax of the predicted probability vector, we obtain an integer corresponding to the model's predicted object class, which we can identify with an object category through the use of this dictionary.
from keras.applications.resnet50 import preprocess_input, decode_predictions
def ResNet50_predict_labels(img_path):
# returns prediction vector for image located at img_path
img = preprocess_input(path_to_tensor(img_path))
return np.argmax(ResNet50_model.predict(img))
Write a Dog Detector¶
While looking at the dictionary, you will notice that the categories corresponding to dogs appear in an uninterrupted sequence and correspond to dictionary keys 151-268, inclusive, to include all categories from 'Chihuahua' to 'Mexican hairless'. Thus, in order to check to see if an image is predicted to contain a dog by the pre-trained ResNet-50 model, we need only check if the ResNet50_predict_labels function above returns a value between 151 and 268 (inclusive).
We use these ideas to complete the dog_detector function below, which returns True if a dog is detected in an image (and False if not).
### returns "True" if a dog is detected in the image stored at img_path
def dog_detector(img_path):
prediction = ResNet50_predict_labels(img_path)
return ((prediction <= 268) & (prediction >= 151))
(IMPLEMENTATION) Assess the Dog Detector¶
Question 3: Use the code cell below to test the performance of your dog_detector function.
- What percentage of the images in
human_files_shorthave a detected dog? - What percentage of the images in
dog_files_shorthave a detected dog?
Answer: 2% of the images in human_files_short have a detected dog, and 100% of the images in dog_files_short have a detected dog.
### Test the performance of the dog_detector function
### on the images in human_files_short and dog_files_short.
n_human_faces = sum([dog_detector(file) for file in human_files_short])
n_dog_faces = sum([dog_detector(file) for file in dog_files_short])
print("Percentage of first 100 human images that have a detected dog: {}%".format(n_human_faces))
print("Percentage of first 100 dog images that have a detected dog: {}%".format(n_dog_faces))
Step 3: Create a CNN to Classify Dog Breeds (from Scratch)¶
Now that we have functions for detecting humans and dogs in images, we need a way to predict breed from images. In this step, you will create a CNN that classifies dog breeds. You must create your CNN from scratch (so, you can't use transfer learning yet!), and you must attain a test accuracy of at least 1%. In Step 5 of this notebook, you will have the opportunity to use transfer learning to create a CNN that attains greatly improved accuracy.
Be careful with adding too many trainable layers! More parameters means longer training, which means you are more likely to need a GPU to accelerate the training process. Thankfully, Keras provides a handy estimate of the time that each epoch is likely to take; you can extrapolate this estimate to figure out how long it will take for your algorithm to train.
We mention that the task of assigning breed to dogs from images is considered exceptionally challenging. To see why, consider that even a human would have great difficulty in distinguishing between a Brittany and a Welsh Springer Spaniel.
| Brittany | Welsh Springer Spaniel |
|---|---|
 |
 |
It is not difficult to find other dog breed pairs with minimal inter-class variation (for instance, Curly-Coated Retrievers and American Water Spaniels).
| Curly-Coated Retriever | American Water Spaniel |
|---|---|
 |
 |
Likewise, recall that labradors come in yellow, chocolate, and black. Your vision-based algorithm will have to conquer this high intra-class variation to determine how to classify all of these different shades as the same breed.
| Yellow Labrador | Chocolate Labrador | Black Labrador |
|---|---|---|
 |
 |
 |
We also mention that random chance presents an exceptionally low bar: setting aside the fact that the classes are slightly imabalanced, a random guess will provide a correct answer roughly 1 in 133 times, which corresponds to an accuracy of less than 1%.
Remember that the practice is far ahead of the theory in deep learning. Experiment with many different architectures, and trust your intuition. And, of course, have fun!
Pre-process the Data¶
We rescale the images by dividing every pixel in every image by 255.
from PIL import ImageFile
ImageFile.LOAD_TRUNCATED_IMAGES = True
# pre-process the data for Keras
train_tensors = paths_to_tensor(train_files).astype('float32')/255
valid_tensors = paths_to_tensor(valid_files).astype('float32')/255
test_tensors = paths_to_tensor(test_files).astype('float32')/255
(IMPLEMENTATION) Model Architecture¶
Create a CNN to classify dog breed. At the end of your code cell block, summarize the layers of your model by executing the line:
model.summary()
We have imported some Python modules to get you started, but feel free to import as many modules as you need. If you end up getting stuck, here's a hint that specifies a model that trains relatively fast on CPU and attains >1% test accuracy in 5 epochs:
Question 4: Outline the steps you took to get to your final CNN architecture and your reasoning at each step. If you chose to use the hinted architecture above, describe why you think that CNN architecture should work well for the image classification task.
Answer: My first step was to review the hinted architecture. I noticed that it achieves a great deal of parameter reduction by using a Global Average Pooling (GAP) layer near the end. I didn't want to do that, because time to complete this task was not going to be extensive on the GPU that I was using. After all, we are only aiming for 1% accuracy with this exercise. I decided to replace the GAP layer with a Flatten layer, which increased the parameter count by a great deal. My reasoning for this was that I hoped the additional parameters would allow me to achieve 2-5% accuracy within ten epochs. I decided to increase the number of filters in each Convolution layer from 8 to 16 to 32, because this technique was suggested in the lessons from the CNN course. I also decided to modify the convolution kernel window sizes to be $3 \times 3$ instead of $2 \times 2$. I do not have a strong intuition that led me to make this modification; it was simply experimentation. I kept the image padding and Max Pooling layers the same as suggested in the hinted architecture. From the few examples I went through in the mini-projects within the CNN course, these choices seemed reasonable. The first time that I fit this model, it ran quite slowly even on the GPU. I was not going to wait for ten epochs to execute. At that point I decided to remove the Convolution layer with 16 filters and the Max Pooling layer immediately after it. This reduced the parameters quite a bit, but it was still several orders of magnitude larger than the hinted architecture. I compiled and fit this model and it executed in a reasonable amount of time, and I was able to achieve 4.4% accuracy.
from keras.layers import Conv2D, MaxPooling2D, GlobalAveragePooling2D
from keras.layers import Dropout, Flatten, Dense
from keras.models import Sequential
model = Sequential()
model.add(Conv2D(filters=8, kernel_size=3, padding='same', activation='relu', input_shape=(224, 224, 3)))
model.add(MaxPooling2D(pool_size=2))
model.add(Conv2D(filters=32, kernel_size=3, padding='same', activation='relu'))
model.add(MaxPooling2D(pool_size=2))
model.add(Flatten())
model.add(Dense(133, activation='softmax'))
model.summary()
Compile the Model¶
model.compile(optimizer='rmsprop', loss='categorical_crossentropy', metrics=['accuracy'])
(IMPLEMENTATION) Train the Model¶
Train your model in the code cell below. Use model checkpointing to save the model that attains the best validation loss.
You are welcome to augment the training data, but this is not a requirement.
from keras.callbacks import ModelCheckpoint
epochs = 10 # the number of epochs used to train the model.
### Do NOT modify the code below this line.
checkpointer = ModelCheckpoint(filepath='saved_models/weights.best.from_scratch.hdf5',
verbose=1, save_best_only=True)
model.fit(train_tensors, train_targets,
validation_data=(valid_tensors, valid_targets),
epochs=epochs, batch_size=20, callbacks=[checkpointer], verbose=1)
Load the Model with the Best Validation Loss¶
model.load_weights('saved_models/weights.best.from_scratch.hdf5')
Test the Model¶
Try out your model on the test dataset of dog images. Ensure that your test accuracy is greater than 1%.
# get index of predicted dog breed for each image in test set
dog_breed_predictions = [np.argmax(model.predict(np.expand_dims(tensor, axis=0))) for tensor in test_tensors]
# report test accuracy
test_accuracy = 100*np.sum(np.array(dog_breed_predictions)==np.argmax(test_targets, axis=1))/len(dog_breed_predictions)
print('Test accuracy: %.4f%%' % test_accuracy)
bottleneck_features = np.load('bottleneck_features/DogVGG16Data.npz')
train_VGG16 = bottleneck_features['train']
valid_VGG16 = bottleneck_features['valid']
test_VGG16 = bottleneck_features['test']
Model Architecture¶
The model uses the the pre-trained VGG-16 model as a fixed feature extractor, where the last convolutional output of VGG-16 is fed as input to our model. We only add a global average pooling layer and a fully connected layer, where the latter contains one node for each dog category and is equipped with a softmax.
VGG16_model = Sequential()
VGG16_model.add(GlobalAveragePooling2D(input_shape=train_VGG16.shape[1:]))
VGG16_model.add(Dense(133, activation='softmax'))
VGG16_model.summary()
Compile the Model¶
VGG16_model.compile(loss='categorical_crossentropy', optimizer='rmsprop', metrics=['accuracy'])
Train the Model¶
checkpointer = ModelCheckpoint(filepath='saved_models/weights.best.VGG16.hdf5',
verbose=1, save_best_only=True)
VGG16_model.fit(train_VGG16, train_targets,
validation_data=(valid_VGG16, valid_targets),
epochs=20, batch_size=20, callbacks=[checkpointer], verbose=2)
Load the Model with the Best Validation Loss¶
VGG16_model.load_weights('saved_models/weights.best.VGG16.hdf5')
Test the Model¶
Now, we can use the CNN to test how well it identifies breed within our test dataset of dog images. We print the test accuracy below.
# get index of predicted dog breed for each image in test set
VGG16_predictions = [np.argmax(VGG16_model.predict(np.expand_dims(feature, axis=0))) for feature in test_VGG16]
# report test accuracy
test_accuracy = 100*np.sum(np.array(VGG16_predictions)==np.argmax(test_targets, axis=1))/len(VGG16_predictions)
print('Test accuracy: %.4f%%' % test_accuracy)
Predict Dog Breed with the Model¶
from extract_bottleneck_features import *
def VGG16_predict_breed(img_path):
# extract bottleneck features
bottleneck_feature = extract_VGG16(path_to_tensor(img_path))
# obtain predicted vector
predicted_vector = VGG16_model.predict(bottleneck_feature)
# return dog breed that is predicted by the model
return dog_names[np.argmax(predicted_vector)]
Step 5: Create a CNN to Classify Dog Breeds (using Transfer Learning)¶
You will now use transfer learning to create a CNN that can identify dog breed from images. Your CNN must attain at least 60% accuracy on the test set.
In Step 4, we used transfer learning to create a CNN using VGG-16 bottleneck features. In this section, you must use the bottleneck features from a different pre-trained model. To make things easier for you, we have pre-computed the features for all of the networks that are currently available in Keras:
- VGG-19 bottleneck features
- ResNet-50 bottleneck features
- Inception bottleneck features
- Xception bottleneck features
The files are encoded as such:
Dog{network}Data.npz
where {network}, in the above filename, can be one of VGG19, Resnet50, InceptionV3, or Xception. Pick one of the above architectures, download the corresponding bottleneck features, and store the downloaded file in the bottleneck_features/ folder in the repository.
(IMPLEMENTATION) Obtain Bottleneck Features¶
In the code block below, extract the bottleneck features corresponding to the train, test, and validation sets by running the following:
bottleneck_features = np.load('bottleneck_features/Dog{network}Data.npz')
train_{network} = bottleneck_features['train']
valid_{network} = bottleneck_features['valid']
test_{network} = bottleneck_features['test']### Obtain bottleneck features from another pre-trained CNN.
bottleneck_features = np.load('bottleneck_features/DogVGG19Data.npz')
train_VGG19 = bottleneck_features['train']
valid_VGG19 = bottleneck_features['valid']
test_VGG19 = bottleneck_features['test']
(IMPLEMENTATION) Model Architecture¶
Create a CNN to classify dog breed. At the end of your code cell block, summarize the layers of your model by executing the line:
<your model's name>.summary()
Question 5: Outline the steps you took to get to your final CNN architecture and your reasoning at each step. Describe why you think the architecture is suitable for the current problem.
Answer: First, I began by downloading the Inception bottleneck features. I then copied the same architecture as used above in the VGG16 bottleneck features and fit the model. My reasoning here was to compute a comparative benchmark for the Inception bottleneck features. The model fitting ran much slower than the VGG16 model and produced worse accuracy. I realized that I could probably achieve a good deal of accuracy improvement by modifying the model architecture, and/or the number of epochs, and/or the batch sizes. However, it would likely require much more time to fit such a model, and I did not think I would be able to achieve at least 60% accuracy in a reasonable amount of epochs. I decided to abandon the Inception bottleneck features.
I then downloaded the VGG19 bottleneck features. Again, I decided to start with the VGG16 model architecture for a comparative benchmark. The performance was great, already above 60%. However, I wanted to modify the architecture and aim for at least 70% accuracy. I noticed that the model was fitting very quickly on the GPU (approximately 1 second per epoch). Obviously this was due to the GAP layer at the beginning of the architecture that was greatly reducing the number of parameters. Unlike above, I decided to keep this GAP layer and not change it to a Flatten layer. In order to gain accuracy improvement I needed to add something, so I decided to add another fully connected Dense layer immediatly after the GAP layer. I chose to put 250 nodes in this layer, with a relu activation. My reasoning for this was simply due to the minimal experience gained in the mini-projects. In those mini-projects, it seems that standard practice is to set the number of nodes in successive Dense layers to be about half of those in preceding layers. Since the output layer has 133 nodes, I figured that 250 nodes in the newly inserted Dense layer would yield robust performance. Also, I decided to add a Dropout layer (with $p=0.25$) between the two Dense layers. I chose this from experience. In several of the mini-projects I tested architectures with and without Dropout layers, and I noticed that a modest amount of weight dropout yielded better validation accuracy when fitting the model. Since the model fitting was running so quickly on the GPU, I decided to run it for 200 epochs with a model checkpointer. My model took about 3 to 4 minutes to execute those 200 epochs, and it achieved over 77% accuracy on the test set.
I believe this architecture is suitable for the current problem because it:
- makes use of existing model weights for a similar problem that were already extensively trained,
- executes extremely fast on a GPU,
- yields satisfactory accuracy results.
Note: My output below only shows output from 20 epochs. I ran the notebook twice after the initial fitting and used the existing saved weights in the saved_models directory each time.
### Define your architecture.
VGG19_model = Sequential()
VGG19_model.add(GlobalAveragePooling2D(input_shape=train_VGG19.shape[1:]))
VGG19_model.add(Dense(250, activation='relu'))
VGG19_model.add(Dropout(0.25))
VGG19_model.add(Dense(133, activation='softmax'))
VGG19_model.summary()
(IMPLEMENTATION) Compile the Model¶
### Compile the model.
VGG19_model.compile(loss='categorical_crossentropy', optimizer='rmsprop', metrics=['accuracy'])
(IMPLEMENTATION) Train the Model¶
Train your model in the code cell below. Use model checkpointing to save the model that attains the best validation loss.
You are welcome to augment the training data, but this is not a requirement.
### Train the model.
checkpointer = ModelCheckpoint(filepath='saved_models/weights.best.VGG19.hdf5',
verbose=1, save_best_only=True)
VGG19_model.fit(train_VGG19, train_targets,
validation_data=(valid_VGG19, valid_targets),
epochs=200, batch_size=20, callbacks=[checkpointer], verbose=2)
(IMPLEMENTATION) Load the Model with the Best Validation Loss¶
### Load the model weights with the best validation loss.
VGG19_model.load_weights('saved_models/weights.best.VGG19.hdf5')
(IMPLEMENTATION) Test the Model¶
Try out your model on the test dataset of dog images. Ensure that your test accuracy is greater than 60%.
### Calculate classification accuracy on the test dataset.
# get index of predicted dog breed for each image in test set
VGG19_predictions = [np.argmax(VGG19_model.predict(np.expand_dims(feature, axis=0))) for feature in test_VGG19]
# report test accuracy
test_accuracy = 100*np.sum(np.array(VGG19_predictions)==np.argmax(test_targets, axis=1))/len(VGG19_predictions)
print('Test accuracy: %.4f%%' % test_accuracy)
(IMPLEMENTATION) Predict Dog Breed with the Model¶
Write a function that takes an image path as input and returns the dog breed (Affenpinscher, Afghan_hound, etc) that is predicted by your model.
Similar to the analogous function in Step 5, your function should have three steps:
- Extract the bottleneck features corresponding to the chosen CNN model.
- Supply the bottleneck features as input to the model to return the predicted vector. Note that the argmax of this prediction vector gives the index of the predicted dog breed.
- Use the
dog_namesarray defined in Step 0 of this notebook to return the corresponding breed.
The functions to extract the bottleneck features can be found in extract_bottleneck_features.py, and they have been imported in an earlier code cell. To obtain the bottleneck features corresponding to your chosen CNN architecture, you need to use the function
extract_{network}
where {network}, in the above filename, should be one of VGG19, Resnet50, InceptionV3, or Xception.
### Write a function that takes a path to an image as input
### and returns the dog breed that is predicted by the model.
def VGG19_predict_breed(img_path):
# extract bottleneck features
bottleneck_feature = extract_VGG19(path_to_tensor(img_path))
# obtain predicted vector
predicted_vector = VGG19_model.predict(bottleneck_feature)
# return dog breed that is predicted by the model
return dog_names[np.argmax(predicted_vector)]
Step 6: Write your Algorithm¶
Write an algorithm that accepts a file path to an image and first determines whether the image contains a human, dog, or neither. Then,
- if a dog is detected in the image, return the predicted breed.
- if a human is detected in the image, return the resembling dog breed.
- if neither is detected in the image, provide output that indicates an error.
You are welcome to write your own functions for detecting humans and dogs in images, but feel free to use the face_detector and dog_detector functions developed above. You are required to use your CNN from Step 5 to predict dog breed.
Some sample output for our algorithm is provided below, but feel free to design your own user experience!
(IMPLEMENTATION) Write your Algorithm¶
def dog_breed_classifier(img_path):
"""
This function accepts a file path to an image and then predicts whether the image has
a human, a dog, or neither in it. If the prediction is 'human' or 'dog', then the
function will predict the dog breed that most closely resembles the image based on
the VGG19 model modification from above (i.e., 'VGG19_model')
"""
isHuman = face_detector(img_path) # boolean: is there a human face in the image?
isDog = dog_detector(img_path) # boolean: is there a dog in the image?
if not isHuman and not isDog:
class_output = 'ImageError: No human or dog in this image.'
breed_output = 'ImageError: Provide an image of a human or dog.'
elif isHuman and not isDog:
class_output = 'This image is a human.'
breed_output = 'Human, you look like a(n) {}.'.format(VGG19_predict_breed(img_path))
elif not isHuman and isDog:
class_output = 'This image is a dog.'
breed_output = 'Dog, you are probably a(n) {}.'.format(VGG19_predict_breed(img_path))
else:
class_output = 'This image might be a human or a dog.'
breed_output = 'Either way, you look like a(n) {}.'.format(VGG19_predict_breed(img_path))
return class_output, breed_output
Step 7: Test Your Algorithm¶
In this section, you will take your new algorithm for a spin! What kind of dog does the algorithm think that you look like? If you have a dog, does it predict your dog's breed accurately? If you have a cat, does it mistakenly think that your cat is a dog?
(IMPLEMENTATION) Test Your Algorithm on Sample Images!¶
Test your algorithm at least six images on your computer. Feel free to use any images you like. Use at least two human and two dog images.
Question 6: Is the output better than you expected :) ? Or worse :( ? Provide at least three possible points of improvement for your algorithm.
Answer: The output is better than expected. The human face_detector() function developed above achieved 80% accuracy with my 15 images. It incorrectly identified three images as humans:
- my dog Molly (a Maltipoo),
- a deer walking through my campsite at the Breaks Interstate Park in Virginia,
- a plate of food (coincidentally venison... unrelated to the deer mentioned above).
I was particularly amused and confused by the mis-detection of the plate of food as a human. The dog_detector() function developed above achieved 100% accuracy with my 15 images. It detected all the dogs and did not mis-label any non-dogs as dogs. These statistics are consistent with the ones we obtained above when we first tested those two functions.
I had eight dog images in my collection, three mixed breeds and five pure breeds. The three mixed breeds are not listed among the 133 dog breed labels, so the best we could hope for with these three would be that the dog breed classifier identified the dog breed as one of the component breeds or a closely related breed. Four of the five pure breeds were identified correctly (80% accuracy). The lone mistake was identifying my dog Maddux (a Rottweiler) as a Beauceron. Note, Beauceron's are often mis-identified as Rottweilers by humans. Of the three mixed breeds I will say that there are 2 successes:
- my dog Buddy (a Cavachon - Bichon Frise & Cavalier King Charles Spaniel) as a Havanese,
- a friend's dog Bailey (a Labrador & Hound mix) as an Italian Greyhound.
A Havanese is a Bichon type breed, so this is a successful breed identification for Buddy. Bailey's true component breed makeup is a little bit of a mystery. He is some kind of Labrador and Hound mix. The lone failure was my dog Molly, who was also identified as a human. I think Molly's image is not ideal; the unusual floor pattern around her could be causing issues. Overall, six of the eight dog breeds (75%) were classified correctly. This is consistent with the model testing done above.
These results were very pleasing considering that the breed classification model was trained without augmented images, and I simply chose 15 unprocessed images from my computer.
Possible improvements for my algorithm:¶
- I could re-design the output of the two detection functions,
face_detector()anddog_detector(), so that in addition to returning a'True'or'False', they also return a probability associated with the detection of a human face or a dog. If I did this, then I could build a new function calleddetection()that used both of those helper functions and returned only'human'or'dog'or'None', depending on the two boolean detection values and the associated probabilities. - I could modify the dog breed classification to work better for mixed breeds. To do this I would have to modify the breed classifier to return a list of dog breeds. It would always return the breed having the highest probability, and it would also append to that list any breed whose probablity exceeded some preselected threshold, e.g., 25%.
- I could design my own
face_detector()anddog_detector()functions using deep learning models of my own architecture, but still perhaps using transfer learning. - I could pre-process the images passed to the classifier by centering detected faces within the image, rotating detected faces so that the eye line is horizontal within the image, and scaling images that appear to be too large or too small within the frame.
paths = glob('test_images/*.jpg')
img_classifications = []
for path in paths:
cls, brd = dog_breed_classifier(path)
img_classifications.append((cls, brd))
## If you want the image outputs to be larger and run vertically sequentially, run this cell.
img_counter = 0
for cls, brd in img_classifications:
print(cls)
img = cv2.imread(paths[img_counter])
cv_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
plt.imshow(cv_rgb)
plt.show()
print(brd)
print('\n\n')
print('=============================================================================')
print('=============================================================================')
print('\n\n')
img_counter += 1
## If you want the image outputs in a subplot figure, run this cell.
fig = plt.figure(figsize=(20, 35))
img_counter = 0
for cls, brd in img_classifications:
ax = fig.add_subplot(5, 3, img_counter+1)
img = cv2.imread(paths[img_counter])
cv_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
ax.imshow(cv_rgb)
ax.set_title(cls, fontsize=14)
plt.xlabel(brd, fontsize=14)
img_counter += 1
